JSON Connector - Kore.ai Docs

Back to Search AI connectors list The JSON connector lets you ingest pre-chunked content directly into Search AI. Content is immediately added to the index, enabling efficient data retrieval. This gives you the flexibility to bring any type of data into the application by structuring it in the required JSON format. Content is uploaded as one or more JSON files, each containing an array of objects where every object represents a chunk. The keys of each object must exactly match the chunk fields defined below.

File Format

Each JSON file must be an array of objects, where each object represents a single chunk. Key requirements:

The file name is used as the recordTitle.
The JSON structure must be an array of objects.
Each object’s keys must match the supported chunk fields listed in the table below.

Field	Description	Mandatory
`chunkText`	Content used to render the final answer for extractive responses and sent to the LLM for generative answers	Yes
`recordUrl`	URL used to generate user references — indicates where the content was originally sourced	Yes
`sourceACL`	List of user identities that have access to this chunk	No
`sourceUrl`	URL of the primary source (e.g., `www.example.org`). If empty, defaults to the value of `recordUrl`	No
`chunkMeta`	Metadata associated with the chunk, used for further processing or embedding generation	No
`chunkTitle`	Title used to render the final answer for extractive responses and sent to the LLM for generative answers	No
`cfa1`	Custom array field available for use based on your requirements	No
`cfa2`	Custom array field available for use based on your requirements	No
`cfa3`	Custom array field available for use based on your requirements	No
`cfa4`	Custom array field available for use based on your requirements	No
`cfa5`	Custom array field available for use based on your requirements	No
`cfs1`	Custom string field available for use based on your requirements	No
`cfs2`	Custom string field available for use based on your requirements	No
`cfs3`	Custom string field available for use based on your requirements	No
`cfs4`	Custom string field available for use based on your requirements	No
`cfs5`	Custom string field available for use based on your requirements	No

Refer to the sample file for a reference implementation.

Configuration Steps

Go to Content > Connectors and add the JSON Connector.
Select the source name to upload the JSON content to. If no source exists, create a new one. Upload the JSON file with your structured data.
Click Proceed. Search AI ingests the content from the uploaded files immediately.
Click the Files tab to view the contents of the connector at any time.

To add more content to an existing source, add the JSON connector again, select the same source name, and upload additional files.

Upload limits:

Up to 10 files per upload
Maximum 15 MB per file
The file name is used as the recordTitle

​File Format

​Configuration Steps

File Format

Configuration Steps